IAI SA7 common mistakes
IAI SA7 · Understand the traps and how to avoid them.
Writing a textbook account of the topic (e.g. a full list of the standard model's assumptions) without connecting a single sentence to the named company, scheme or portfolio in the question.
Under time pressure, recall of memorised notes is faster and feels safer than the harder task of selecting the two or three concepts that actually bite in this scenario; candidates also misread the scenario as decoration rather than the source of marks.
What to do instead
Before writing, spend 60-90 seconds converting the scenario into a one-line list of facts (entity, objective, constraint, time horizon, jurisdiction). Then force every paragraph to name a fact from that list: 'For [entity], the binding constraint is X, so the relevant assumption is Y because Z.' Practise by re-answering past questions with the scenario facts highlighted and count the paragraphs that never mention one.
Answering 'discuss' or 'evaluate' with a balanced list of pros and cons that stops before any judgement, or answering 'recommend' with a description of options and no chosen option.
Lists are easy to generate and feel productive; committing to a view carries the risk of being wrong, and candidates do not realise that the marks for synthesis sit in the final judgement rather than in the enumeration.
What to do instead
Underline the command verb and rewrite it as a deliverable in the margin ('evaluate -> a reasoned verdict with the criteria stated'). Use the structure: criteria -> option-by-option assessment against those criteria -> explicit recommendation -> conditions and risks of that recommendation. If the answer does not contain the words 'I would recommend...', it is incomplete.
Spending 20-25 minutes on the opening 5- or 6-mark part because it feels familiar, then writing two rushed sentences for a 12-mark part at the end.
The first question is usually on core material the candidate knows well and writing about it is psychologically rewarding; there is no clock discipline because the paper has no per-question time signal.
What to do instead
Impose a marks-based budget before starting: roughly 1.8 minutes per mark for a three-hour paper, and write the target finish time for each part in the margin. When the target passes, stop mid-sentence, leave a gap and move on. Rehearse this in timed mocks, not untimed practice, and review which parts were habitually under-answered.
Presenting a single final figure (e.g. a projected fund value or a capital requirement) with no formula, no stated inputs, no units and no comment on whether the number is plausible.
Candidates treat the calculation as the answer when marks are awarded for the method, the assumptions and the interpretation; spreadsheet-era habits of quoting only outputs reinforce this.
What to do instead
Write the answer as a short audit trail: formula in symbols, values substituted, intermediate result, units and rounding, then one sentence interpreting the result and one sanity check (order of magnitude, direction versus a known benchmark). If an assumption is not given in the question, state it explicitly rather than silently choosing it.
Ignoring data supplied in the question and substituting remembered market figures, or omitting an assumption given in the stem (discount rate, inflation, mortality basis) so the answer is internally inconsistent.
Candidates trust remembered numbers more than unfamiliar question data, and the 15 minutes of reading time is used to read the questions rather than to inventory the facts and figures they contain.
What to do instead
Use the reading time to build a table of every number and stated assumption in the paper, with the part it belongs to. In the answer, quote the figure you are using before you use it. If you deliberately depart from a given assumption, say so and justify it in one sentence.
Proposing a single solution where the question asks for alternative approaches, or assessing options against the candidate's own preferences rather than the objectives stated in the scenario.
Candidates default to the 'textbook' answer for the product or problem type and do not extract the decision criteria the question actually supplies, so the evaluation is untethered from what the client cares about.
What to do instead
Write the decision criteria from the scenario first (cost, risk appetite, liquidity, regulatory constraint, timescale), then produce at least three genuinely different options, including doing nothing or phasing the change, and score each against the same criteria in a compact table with a short narrative under each row.
Identifying an ethical or professional conflict but stopping at 'this should be escalated', without naming the principle at stake, the person to whom it should be escalated, the advice to give, or the record to keep.
Candidates treat the professional guidance as background reading rather than as an answer template, so they produce a moral reaction instead of a process a real actuary would follow.
What to do instead
Adopt a fixed five-step template and use it every time: identify the issue and the parties; identify the relevant principle (integrity, competence and care, impartiality, compliance, openness); state what advice you would give and to whom; state the escalation route and the point at which you would refuse to sign; state what you would document. One line per step earns more than a paragraph of concern.
Answering a regulatory or tax question with the rules of the candidate's home market when the scenario is set elsewhere, or asserting specific local rules with false confidence.
Candidates revise the regime they work in, and question setters deliberately place scenarios in other jurisdictions to test principle-based reasoning rather than rule recall.
What to do instead
Open with an explicit jurisdiction assumption ('assuming the scheme is governed by X, the position is...; if instead Y applies, the effect reverses because...'). Where the local detail is genuinely unknown, reason from the purpose of the regulation, state the direction of the effect on the client and flag that local advice is needed. Never present an uncertain rule as fact.
Presenting projections, valuations or capital figures as single deterministic numbers with no sensitivity, no range and no statement of the key risks and how they would be mitigated.
Candidates equate a correct central calculation with a complete answer, and sensitivity analysis is the first thing dropped when time is short because it is written last.
What to do instead
Reserve the final two or three sentences of every numerical part for uncertainty: name the two or three assumptions that dominate the result, state the direction (and where possible the rough magnitude) of the effect of a change in each, and say what the client could do about it. Write this section before polishing any other part of the calculation.
Submitting a wall of unbroken prose with no headings, no signposting and no separation between sub-parts, so the marker cannot locate the points being made.
Handwriting speed or typing under time pressure pushes candidates into continuous prose, and candidates assume the marker will read the whole answer carefully rather than mark against a scheme.
What to do instead
Adopt a fixed layout: the sub-part number as a heading, one short sentence stating the approach, bullet points with one idea per bullet, and a bold or underlined one-line conclusion. Leave a blank line between sub-parts. Practise on past papers with a marker's eye: if a stranger could not tick six separate points in ten seconds, restructure.